Amazon will give you essentially a dedicated piece of hardware, albeit with your instance running in para-virtualization. http://aws.amazon.com/ec2/hpc-applications/ Features: -10GB interconnect between cluster compute instance types (FULL, unshared 10GB if they run in cluster mode) -1690 GB of local, bare-metal i/o storage (real storage- not ephemeral) -same access to EBS Volumes as regular instances -23 GB of memory -CentOS 5.4 I'm planning GlusterFS on EC2 testing this summer, and am curious what folks think of Amazon's new compute clusters? In a replicated setup, everyone would love the 10GB interconnects and the very fast ~1.5TB available (with more storage available via LVM and EBS), but what can one do with that much memory? When would GlusterFS derive additional benefits from that? Jack Murgia Owner, Senior Network Engineer Cloud Controllers Amazon Web Services Solution Provider (415) 494-2855 Phone http://www.cloudcontrollers.com/